WEBVTT - Monologue: How The Media Keeps Inflating Bubbles 0:00:02.520 --> 0:00:06.640 Media. Hello, and welcome to this week's Better Offline Monologue. 0:00:06.640 --> 0:00:16.320 I am your host ed Zeitron, of course, So if 0:00:16.320 --> 0:00:18.279 you're listening to this as the episode's are, you're in 0:00:18.280 --> 0:00:20.240 the middle of a two part episode about the systemic 0:00:20.320 --> 0:00:23.400 risk caused by Open Ai, a company that's always been 0:00:23.480 --> 0:00:27.080 unsustainable and is ultimately rigged for collapse. But today I 0:00:27.080 --> 0:00:28.960 wanted to turn to a question I've been asked a 0:00:28.960 --> 0:00:31.680 few times, what are the ways the media can avoid 0:00:31.760 --> 0:00:33.560 doing this in the future, and what are the things 0:00:33.560 --> 0:00:36.920 that they could have done differently? Well, it starts pretty simple. 0:00:37.360 --> 0:00:39.639 I don't believe the media, and this partly falls upon 0:00:39.680 --> 0:00:42.479 the people running media outlets actually knows enough about the 0:00:42.479 --> 0:00:44.839 subject matter, be it the technical side of the financial 0:00:44.840 --> 0:00:47.520 side of these companies. If you've ever read an article 0:00:47.560 --> 0:00:50.360 about tech that just didn't seem to make sense, like 0:00:50.520 --> 0:00:53.239 say about an AI company, like an obtuse series of 0:00:53.280 --> 0:00:56.240 sentences that sounds rational, but when you really think about 0:00:56.240 --> 0:00:58.440 it doesn't explain what it does. It's because the writer 0:00:58.560 --> 0:01:01.600 doesn't actually understand what this saying. And that's because they're 0:01:01.600 --> 0:01:04.080 really given the time or incentivized in any way for 0:01:04.240 --> 0:01:06.800 knowing what it is they're talking about they're just there 0:01:06.840 --> 0:01:09.520 to kind of get it out the door, and well 0:01:09.680 --> 0:01:11.920 that an unalarming amount of tech writers are edited by 0:01:12.000 --> 0:01:15.600 editors that don't know a fucking thing. When chat GPT launched, 0:01:15.600 --> 0:01:18.319 the press absolutely lost their shit, despite nobody being able 0:01:18.319 --> 0:01:20.160 to describe what it actually did and why it was 0:01:20.160 --> 0:01:22.840 the future other than they could create an image or 0:01:22.880 --> 0:01:25.160 a block of text based on a prompt, and this 0:01:25.280 --> 0:01:27.360 was about the level of nuance that we'd see applied 0:01:27.400 --> 0:01:29.920 to open AI for pretty much the rest of history. 0:01:30.400 --> 0:01:33.199 Egregious to extrapolations were made in part because the media 0:01:33.319 --> 0:01:35.440 was far too willing to just copypaste or quote whatever 0:01:35.440 --> 0:01:38.240 Sam Altman said. When interviewed on stage at the Wall 0:01:38.280 --> 0:01:41.119 Street Journals Tech Live event in October twenty twenty three, 0:01:41.240 --> 0:01:43.720 a reporter allowed Sam Moltman to say that chat GPT 0:01:43.840 --> 0:01:47.040 and I quote has this laval reasoning capacity that's going 0:01:47.040 --> 0:01:49.760 to get better and better, and this was a great 0:01:49.800 --> 0:01:52.120 moment to say, I'm sorry, Sam, what the fuck are 0:01:52.160 --> 0:01:54.240 you talking about? And to be clear, this was October 0:01:54.240 --> 0:01:56.760 twenty twenty three. It would be just under a year 0:01:57.360 --> 0:02:00.240 later that they would release an actual reasoning product. No 0:02:00.280 --> 0:02:03.080 one's bothered to return to that of course, I realize 0:02:03.120 --> 0:02:06.480 I sound like I'm being nasty or facetious. But here, right, 0:02:06.560 --> 0:02:08.919 here was an opportunity to push back on the narrative, 0:02:09.200 --> 0:02:11.440 in part because the narrative was a guy will say 0:02:11.520 --> 0:02:15.040 fucking anything and anybody will print him. The question, of course, 0:02:15.080 --> 0:02:17.360 would have been can you explain what this means? And 0:02:17.680 --> 0:02:20.120 once he mumbles out some nonsense, say I'm still not 0:02:20.160 --> 0:02:23.240 sure I understand. I genuinely believe that there are some 0:02:23.280 --> 0:02:25.640 reporters who don't want to push against the grain, but 0:02:25.680 --> 0:02:28.040 I think that there are plenty more that are scared 0:02:28.080 --> 0:02:31.200 of being wrong. Well, you're still wrong if you allow 0:02:31.240 --> 0:02:34.400 a man to lie. Sam Mortman became a billionaire because 0:02:34.440 --> 0:02:37.360 he was able to blather on nonsensically about whatever he wanted, 0:02:37.480 --> 0:02:40.440 knowing that the crowd of Gordoner's business types would rather say, Wow, 0:02:40.480 --> 0:02:42.960 he's so smart, I asked the future, than admit that 0:02:42.960 --> 0:02:46.200 they have no idea what he fucking meant. Really, fixing 0:02:46.240 --> 0:02:49.120 these problems starts with making people like Sam Mortman of 0:02:49.120 --> 0:02:53.160 Open AI and Dario Amadeo of Anthropic actually explain themselves 0:02:53.280 --> 0:02:55.760 and holding their feet to the fire even lightly on 0:02:55.800 --> 0:02:58.800 their promises, and the economics underlying their companies. You can 0:02:58.880 --> 0:03:01.840 beat this shit, you can do it. And early on 0:03:02.240 --> 0:03:04.760 Sam Mortman actually been pushed back on. He would not 0:03:04.880 --> 0:03:07.400 have been able to do this. And indeed, had everyone 0:03:07.440 --> 0:03:10.560 not just copypasted, anytime Sam Mortman said anything about, oh, 0:03:10.560 --> 0:03:13.480 I don't know this chat GPT will eventually be an 0:03:13.520 --> 0:03:15.680 intelligent friend and knows everything about you, I believe he 0:03:15.760 --> 0:03:18.720 said that, or he just said AI will be These 0:03:18.760 --> 0:03:20.800 are the times to say I don't know print it 0:03:20.800 --> 0:03:24.000 in say Sam Mortman lies. Sam Morman made something up. 0:03:24.040 --> 0:03:26.440 I realized the media writ large is very bad at this, 0:03:26.520 --> 0:03:30.400 but it's so much easier in tech. The fact that 0:03:30.440 --> 0:03:33.280 both Open AI and Anthropic burned over five billion dollars 0:03:33.280 --> 0:03:35.920 in twenty twenty four is abominable and any and all 0:03:35.960 --> 0:03:38.200 interviews should have brought this up and doggedly demanded a 0:03:38.240 --> 0:03:41.840 timeline for profitability and not accepted vagueness or dodging. And 0:03:41.880 --> 0:03:44.640 I just want to be clear, I know that this 0:03:44.840 --> 0:03:46.720 if you remember the media listening to so like, I 0:03:46.720 --> 0:03:48.520 won't get Sam Mormon again, I won't get I won't 0:03:48.520 --> 0:03:51.720 get dariy Amaday, who fucking cares? These interviews suck every 0:03:52.080 --> 0:03:54.920 I think I've listened to or watched every Sam Mortman 0:03:54.960 --> 0:03:59.760 interview that's online, and they're boring. They're like eating cardboard. 0:04:00.040 --> 0:04:03.280 He's not an endearing speaker. Warrio Ama days even worse 0:04:03.320 --> 0:04:07.320 that bloviating fuck where wah blah Blind twenty twenty seven. 0:04:07.560 --> 0:04:09.640 The computer is going to be my best friend and 0:04:09.680 --> 0:04:12.960 my girlfriend. Well he's married. I'm not saying Dario's anyway. 0:04:13.120 --> 0:04:15.040 Daria's not going to fuck the computer. You've got me 0:04:15.080 --> 0:04:18.880 on the record. The boint is you're afraid of losing access, 0:04:18.880 --> 0:04:21.320 but you're also afraid of having a hostile interview. First 0:04:21.360 --> 0:04:24.799 of all, skill issue. Second of all, why are you afraid? 0:04:25.400 --> 0:04:28.080 Why would you possibly be afraid? These men are far 0:04:28.120 --> 0:04:31.080 from invincible. Sam Wrtman is a solid conn artist, stand 0:04:31.080 --> 0:04:33.599 a carnival barker, but when faced with blunt questions and 0:04:33.640 --> 0:04:36.679 even the slightest hint of memory about what he just said, 0:04:36.760 --> 0:04:40.279 he'd crumble. All of them would honestly Imdia train people 0:04:40.320 --> 0:04:43.480 for a living, these people a week. They're poorly trained 0:04:43.480 --> 0:04:45.680 and speciously informed and have no idea what to do. 0:04:45.839 --> 0:04:48.839 If you just refuse to black like you just sit 0:04:48.920 --> 0:04:50.960 on a question, if you're just like no, I'm not 0:04:51.080 --> 0:04:55.159 happy with your answer, I don't understand that doesn't make sense, 0:04:55.720 --> 0:04:59.680 or even hey, why won't you give me a straight answer? 0:05:00.320 --> 0:05:03.720 Crazy crazy question? You could just ask any of them there, 0:05:03.839 --> 0:05:06.680 You could do it today now. The other problem is 0:05:06.720 --> 0:05:09.919 the journalists too regularly find ways to ask these obtuse, 0:05:10.040 --> 0:05:12.039 muddied questions, in part because they want to find a 0:05:12.040 --> 0:05:14.039 way to sound like they're being aggressive to their readers 0:05:14.240 --> 0:05:16.479 without ever really showing any real aggression. And let me 0:05:16.480 --> 0:05:19.240 give you one of my favorite examples. In May twenty 0:05:19.240 --> 0:05:21.920 twenty four, Neelai Patel of The Verge, a man who 0:05:21.960 --> 0:05:23.960 has never met an executive outside of into it that 0:05:24.000 --> 0:05:26.640 he didn't want to let ramble, interviewed sund up As Shy, 0:05:26.720 --> 0:05:28.680 the CEO of Google, and gave one of the single 0:05:28.680 --> 0:05:32.240 worst interviews I've seen in my life. Neili Patel, as 0:05:32.279 --> 0:05:34.799 a lawyer, as the editor in chief of The Virgin, 0:05:34.839 --> 0:05:37.640 a reporter with over a decade of experience possibly more, 0:05:38.279 --> 0:05:42.120 asked multiple questions of over a hundred words objection compound 0:05:42.160 --> 0:05:44.960 question which may seem like he's being thoughtful, but is 0:05:45.000 --> 0:05:47.320 actually what I like to call the buffet, an attempt 0:05:47.320 --> 0:05:50.000 to give the illusion of nuanced analysis and conversation where 0:05:50.240 --> 0:05:52.760 you're actually giving the person you're interviewing the opportunity to 0:05:52.760 --> 0:05:55.600 answer the question they'd like to. Sometimes these points are 0:05:55.640 --> 0:05:58.440 more like rambling statements, which again may seem harmless, like 0:05:58.520 --> 0:06:00.719 Neelai is trying to have a conversation, but this is 0:06:00.760 --> 0:06:04.240 the CEO of fucking Google. You're interviewing him, not up. 0:06:04.800 --> 0:06:07.840 At one point, Nelai spends one hundred and fifty nine 0:06:07.920 --> 0:06:11.479 words asking Sun dar Pashai whether he expected publishers to 0:06:11.520 --> 0:06:14.880 act negatively to AI powered search results, but he did 0:06:14.920 --> 0:06:17.200 so in such a cludgy way that Sun Dapashai is 0:06:17.200 --> 0:06:19.600 able to wave it off, at which point Neeli commits 0:06:19.640 --> 0:06:22.000 the ultimate sin one very common in the tech media, 0:06:22.080 --> 0:06:24.360 where he doesn't listen to Peshai's answer and immediately asks 0:06:24.400 --> 0:06:27.520 another question based on some theory he has called Google zero, 0:06:27.600 --> 0:06:32.720 whether traffic from Google it ends, which was already happening. Neelike, 0:06:32.760 --> 0:06:34.760 you talk to one of the many talented people you're 0:06:34.800 --> 0:06:38.640 fucking outlet. This, by the way, is the real fundamental 0:06:38.640 --> 0:06:41.520 flaw of American journalism, where an interviewer asks the question, 0:06:41.640 --> 0:06:44.159 doesn't listen to the answer, then asks another I cannot 0:06:44.200 --> 0:06:47.440 express enough how many times this happens with tech CEOs, 0:06:47.560 --> 0:06:49.839 how much advantage they take of it and how quickly 0:06:49.839 --> 0:06:51.839 they collapse. If you listen to their answer, and I 0:06:51.880 --> 0:06:54.120 don't know, ask the follow up question relevant to it. 0:06:54.560 --> 0:06:57.640 Nelai's interview could have been far more direct. Hey, why 0:06:57.680 --> 0:07:00.479 does Google search suck? Now? I've got eight different examples 0:07:00.520 --> 0:07:03.160 of it sucking? Why are they so bad? Neeli does 0:07:03.160 --> 0:07:05.720 have an example searching for best chromebook, and he has 0:07:05.760 --> 0:07:10.000 this clumsy, mumbling, meandering bullshit thing about how in the query, well, 0:07:10.040 --> 0:07:12.800 the query in there. I didn't say what is the 0:07:12.800 --> 0:07:14.840 best chromebook? I did this and that, and he just 0:07:14.920 --> 0:07:19.840 like fumbles around with it and then allows allows some 0:07:19.960 --> 0:07:22.680 doctor to just kind of ramble him. Neli, you could 0:07:22.760 --> 0:07:25.160 just ask a simple fucking question. You're a lawyer, man, 0:07:25.640 --> 0:07:27.680 I don't know. I don't know, man. Maybe you're more 0:07:27.680 --> 0:07:30.640 concerned being fucking famous. The reason I'm so angry is 0:07:30.640 --> 0:07:33.000 that the AI bubble was inflated on interviews like these. 0:07:33.120 --> 0:07:35.520 I realized Neeli Patel wants to continue getting access to 0:07:35.520 --> 0:07:38.720 the executives, But how do he directly said search is bad, 0:07:38.760 --> 0:07:40.680 here's the proof it's bad, and AI results are worse, 0:07:40.800 --> 0:07:42.560 and refuse to back off of it. That would have 0:07:42.640 --> 0:07:45.280 had a meaningful effect on Google's willingness to push this shit. 0:07:45.800 --> 0:07:49.000 It would have shown other journalist's solidarity that they too 0:07:49.480 --> 0:07:51.800 could step up and spit in the face of these people. 0:07:51.960 --> 0:07:53.720 And I know some of you are going to say, heed, 0:07:53.840 --> 0:07:56.320 you can't spit in the face of them. I only 0:07:56.320 --> 0:07:58.280 feel like that because they want to run away so bad, 0:07:58.360 --> 0:08:00.760 because they don't want to have a real interview, because 0:08:00.760 --> 0:08:02.880 they don't have anything to say, they don't want to 0:08:02.920 --> 0:08:05.520 say it. And the same goes for Samultman or Warrio 0:08:05.680 --> 0:08:09.280 Dario ama Day or any other AI executive. They've garnered 0:08:09.320 --> 0:08:11.760 thousands of headlines and billions of dollars of funding by 0:08:11.800 --> 0:08:15.360 making shit up, or using lives of emission, or just 0:08:15.400 --> 0:08:17.520 being very vague and allowing people to fill in the 0:08:17.560 --> 0:08:19.880 gaps for them, which then gets published in the media, 0:08:19.880 --> 0:08:21.960 which then gets pushed to investors who then line up 0:08:21.960 --> 0:08:24.360 to invest in shit that doesn't work, or does not 0:08:24.520 --> 0:08:28.360 exist or can never exist. I even believe that access 0:08:28.440 --> 0:08:31.120 journalism can continue. We just need to show that there's 0:08:31.120 --> 0:08:34.640 a deep intolerance across the industry for marketing bullshit. A 0:08:34.640 --> 0:08:36.720 brick wall in front of anyone who would bullshit and 0:08:36.760 --> 0:08:40.240 an interest in having actual conversations versus regurgitated talking points 0:08:40.320 --> 0:08:43.280 and nodding to an audience of half conscious patagonia gargoyles. 0:08:43.840 --> 0:08:45.520 And I want to bring something up. I run a 0:08:45.520 --> 0:08:48.680 PR firm, immedia trained people. This is actually how fucking 0:08:48.760 --> 0:08:51.679 startups are talked to by journalists. You're a seed startup 0:08:51.720 --> 0:08:54.079 or a series A. You get asked these fucking questions 0:08:54.120 --> 0:08:57.680 all the time. You get grilled, you get absolutely beaten 0:08:57.720 --> 0:09:00.480 to shit. But once you reach a certain scale, once 0:09:00.520 --> 0:09:03.720 you're worth i don't know, some three hundred billion theoretical 0:09:03.800 --> 0:09:06.720 dollars that you fit into a narrative, suddenly the gloves 0:09:06.720 --> 0:09:11.320 are off and that's when the jerkins starts. I realized 0:09:11.360 --> 0:09:14.400 that was quite gross. But look, if you're listening to 0:09:14.400 --> 0:09:15.839 this as a member of the media and think I'm 0:09:15.880 --> 0:09:19.000 being mean to these executives, there's absolutely nothing stopping you 0:09:19.040 --> 0:09:22.600 from having fun, interesting conversations about the shit you love. 0:09:22.920 --> 0:09:26.480 There's nothing, nothing, nothing, nothing wrong with knowing subjects deeply 0:09:26.520 --> 0:09:29.800 and finding reasons to be excited. But please, fuck actually 0:09:29.880 --> 0:09:32.720 learn what you're talking about and be excited about something 0:09:32.760 --> 0:09:36.640 that's actually happening. But I think that that might be 0:09:36.640 --> 0:09:38.800 the problem deep down. That's the art and more problem 0:09:38.800 --> 0:09:41.320 with the tech media. I don't think enough people writing 0:09:41.320 --> 0:09:44.520 about technology actually know or give a shit about technology, 0:09:44.800 --> 0:09:48.280 or at least know enough to do their jobs. They 0:09:48.280 --> 0:09:51.520 see these executives as their sports teams factions to ally behind, 0:09:51.720 --> 0:09:53.400 only learning enough about the take to be able to 0:09:53.400 --> 0:09:55.760 write embargoed articles about whatever the next thing that open 0:09:55.760 --> 0:09:59.640 aiye does, even if they don't really understand or care readers. 0:10:00.000 --> 0:10:02.800 Listners deserved better, and when all of this shit collapses, 0:10:02.800 --> 0:10:04.520 I believe they'll start treating some of the members of 0:10:04.520 --> 0:10:06.480 the tech media with well deserved scorn.